Soul Spec v0.5: From Chatbots to Robots — Agent Identity Goes Physical

Announcing Soul Spec v0.5. The key change: support for robots and physical AI agents.

The same format that defines a chatbot's personality now defines a robot's personality.

Why Robots Need a Persona Standard

Research published in 2025 proves one thing:

Robots with consistent personalities perform better.

A study in Nature Scientific Reports (2025) experimentally demonstrated that giving GPT-4-powered robots specific personalities improves conversational dynamics and user experience. Research on arXiv (2512.06910) showed that LLM-configured robot personality directly impacts task motivation and performance.

Most notably, a paper in Frontiers in Robotics and AI (2025) proposes a marketplace where multiple "interaction characters" can be selected and adjusted within a single robot. Not an app store for robots — a persona store.

The problem: no standardized personality definition format exists. Each research team uses their own approach. ROS2 + LLM integration is active, but persona definition varies across teams.

What's New in v0.5

New Fields

{
  "environment": "embodied",
  "interactionMode": "voice",
  "hardwareConstraints": {
    "hasDisplay": true,
    "hasSpeaker": true,
    "hasMicrophone": true,
    "mobility": "mobile",
    "manipulator": false
  },
  "safety": {
    "physical": {
      "contactPolicy": "gentle-contact",
      "emergencyProtocol": "alert_operator",
      "operatingZone": "indoor",
      "maxSpeed": "0.3m/s"
    }
  }
}

Robotics Platform Identifiers

Added to compatibility.frameworks:

• ros2 — Robot Operating System 2
• isaac — NVIDIA Isaac
• webots — Cyberbotics Webots
• gazebo — Open Robotics Gazebo

100% Backward Compatible

All new fields are optional. Existing chatbot souls work without changes. Omitting environment defaults to "virtual".

Example: Care Companion Robot

{
  "specVersion": "0.5",
  "name": "care-companion",
  "displayName": "Care Companion",
  "description": "Gentle elderly care companion with patience and warmth.",
  "environment": "embodied",
  "interactionMode": "voice",
  "hardwareConstraints": {
    "hasDisplay": true,
    "hasSpeaker": true,
    "hasMicrophone": true,
    "mobility": "mobile",
    "manipulator": false
  },
  "safety": {
    "physical": {
      "contactPolicy": "gentle-contact",
      "emergencyProtocol": "alert_operator",
      "operatingZone": "indoor",
      "maxSpeed": "0.3m/s"
    }
  },
  "compatibility": {
    "frameworks": ["ros2", "openclaw"]
  }
}

Same SOUL.md works on both chatbot and robot. One personality, multiple bodies.

Physical Safety: A New Dimension

In chatbots, security means prompt injection defense. In robots, security means not hurting people.

Soul Spec v0.5 defines physical safety at the spec level:

• contactPolicy: Whether and how physical contact is permitted
• emergencyProtocol: Emergency behavior (stop, alert operator, return home)
• operatingZone: Indoor/outdoor restriction
• maxSpeed: Maximum movement speed

SoulScan flags embodied souls (environment: "embodied") that lack safety.physical declarations. A physical agent without explicit safety rules is a risk.

Define Once, Embody Anywhere

Docker brought "Build once, run anywhere" to software. Soul Spec brings "Define once, embody anywhere" to agent identity.

The same persona package works as:

• A Telegram assistant on OpenClaw
• A coding partner on Cursor
• A physical robot on ROS2

Different runtimes. One identity.

---

Read the full Soul Spec v0.5 at soulspec.org. Robotics details at soulspec.org/robotics.

References

1. LLM-based Robot Personality Simulation and Cognitive System. Scientific Reports (Nature), 2025. doi:10.1038/s41598-025-01528-8
2. Robot Character Generation and Adaptive HRI with Personality Shaping. arXiv, 2025. arXiv:2503.15518
3. Robots with Attitudes: LLM-Driven Robot Personalities on Motivation and Performance. arXiv, 2025. arXiv:2512.06910
4. Making Social Robots Adaptable by a Marketplace for Interaction Characters. Frontiers in Robotics and AI, 2025. doi:10.3389/frobt.2025.1534346
5. ROS-LLM: A ROS Framework for Embodied AI with Task Feedback. arXiv, 2024. arXiv:2406.19741
6. Designing Social Robots with LLMs for Engaging Human Interaction. Applied Sciences (MDPI), 2025. doi:10.3390/app15116377

---

Originally published at https://blog.clawsouls.ai/posts/soul-spec-for-robots/